Using artificial neural networks for identification of electrical losses in transformers during the manufacturing phase

Abstract

The paper describes a novel neural model to estimate electrical losses in transformer during the manufacturing phase. The network acts as an identifier of structural features on electrical loss process, so that output parameters can be estimated and generalized from an input parameter set. The model was trained and assessed through experimental data taking into account core losses, copper losses, resistance, current and temperature. The results obtained in the simulations have shown that the developed technique can be used as an alternative tool to make the analysis of electrical losses on distribution transformer more appropriate regarding to manufacturing process. Thus, this research has led to an improvement on the rational use of energy.